Real-Time Implementation of Long-Horizon Direct Model Predictive Control on an Embedded System

Abstract

This paper deals with the real-time implementation of a long-horizon finite control set model predictive control (FCS-MPC) algorithm on an embedded system. The targeted application is a medium-voltage drive system which means that operation at a very low switching frequency is needed so that the switching power losses are kept relatively low. However, a small sampling interval is required to achieve a fine granularity of switching, and thus ensure superior system performance. This renders the real-time implementation of the controller challenging. To facilitate this, a high level synthesis (HLS) tool, which synthesizes C++ code into VHDL, is employed to enable a higher level of abstraction and faster prototype development of the real-time solver of the long-horizon FCS-MPC problem, namely the sphere decoder. Experimental results based on a small-scale prototype, consisting of a three-level neutral point clamped (NPC) inverter and an induction machine, confirm that the algorithm can be executed in real time within the targeted control period of 25 $\mu$s.